Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Gorilla: Large Language Model connected with massive APIs

gorilla.cs.berkeley.edu

Post reply on HN
Print view View original

61–70 of 123 posts

Re: Gorilla: Large Language Model connected with massive APIs

#61

Post by telotortium »

ImHereToVote wrote: post #48
ethanbond wrote: post #41

Earlier quoted context omitted.

Sure, my argument is that there is zero evidence whatsoever we will be able to prevent these from becoming dangerous or that we’d be able to stop deployment once they do. All technologies are dangerous, and many of the most dangerous ones correctly have tons and tons of safeguards around them both as intrinsic properties of the technology (e.g. it takes nationstate resources to produce a nuke) and extrinsic constrain…

I wish I could create an AGI that would be able to create undetectable bots to upvote this comment hundreds of thousands of times.

And I likewise, but to downvote instead.

Re: Gorilla: Large Language Model connected with massive APIs

#62

Post by pixl97 »

bioemerl wrote: post #54
pixl97 wrote: post #45

Earlier quoted context omitted.

Sorry, was busy plugging my paperclip optimizer into the internet, what's up exactly?

Do you have a paperclip optimizer? No such thing exists, it's science fiction

I mean corporations are the weak versions of such optimizers already.

DNA/RNA is a strong version, but luckily we're aligned with it by evolution, but still at risk of extinction by disease.

Re: Gorilla: Large Language Model connected with massive APIs

#63

Post by pixl97 »

killingtime74 wrote: post #52
pixl97 wrote: post #44

Earlier quoted context omitted.

>History Right, I learned from Napoleon about the dangers of AI... The large land mammals were doing pretty well too, until a bipedal superintelligence showed up. History is a get thing to learn from and a very poor thing to assume is going to repeat.

You've just provided evidence that humans are the most dangerous, not anything else on this planet.

I mean, so far yes. At the same time we have tons of people spending billions of dollars to make sure that's not the case, and I'm pretty sure they are going to succeed.

Re: Gorilla: Large Language Model connected with massive APIs

#64

Post by bioemerl »

soperj wrote: post #59
bioemerl wrote: post #53

Earlier quoted context omitted.

You would have more weight behind your words if there was any sort of precedent for AI as it exists today being super intelligent, or that you could make a case for some method by which you could suddenly ramp out of control without our notice. But as is, all you're doing is making theories, and there's not much of any precedent for any of them.

> that you could make a case for some method by which you could suddenly ramp out of control without our notice. lots of things ramp up without our notice. Like CO2 in the atmosphere, or Ozone depleting chemicals, or PFAS in the water, microplastics in the ocean. I don't think it's a small list.

CO2 has not ramped up without notice, nor did ozone nor pfas. There are active efforts too handle each of them, and the only reason we aren't cracking down on most of those harder is that the utility they provide to human beings is far greater than the damages right now.

Re: Gorilla: Large Language Model connected with massive APIs

#65

Post by generalspecific »

robwwilliams wrote: post #55
petters wrote: post #36

Earlier quoted context omitted.

> Maybe this is a good cause to reassess the premise of alignment as a valuable goal? Could you elaborate here? Alignment seems pretty obviously a good thing.

Alignment assumes a well agreed foundational philosophy on what is good, what is fair, what is doable today and tomorrow. Yes, HN contributors might have shared goals for AGI alignment—but we are not the world—-we are a thin slice of one culture.

I think a more individualistic definition of alignment could say that an AI that a person is directing doesn't do something that person does not desire - this definition removes the "foundational philosophy of what is good" problem, but does leave the "lunatic wants to destroy the world with AIs help" problem. Tricky times ahead

Re: Gorilla: Large Language Model connected with massive APIs

#66

Post by bioemerl »

pixl97 wrote: post #62
bioemerl wrote: post #54

Earlier quoted context omitted.

Do you have a paperclip optimizer? No such thing exists, it's science fiction

I mean corporations are the weak versions of such optimizers already. DNA/RNA is a strong version, but luckily we're aligned with it by evolution, but still at risk of extinction by disease.

I'm sorry, but this is nonsensical. These little AI have trouble putting more than a couple of strings of sentences together without going incoherent.

The fact that corporations try to improve themselves and innovate with time has nothing to do with AI or LLMs and their risks.

The fact that evolution exists also has nothing to do with it.

Re: Gorilla: Large Language Model connected with massive APIs

#67

Post by soperj »

bioemerl wrote: post #64
soperj wrote: post #59

Earlier quoted context omitted.

> that you could make a case for some method by which you could suddenly ramp out of control without our notice. lots of things ramp up without our notice. Like CO2 in the atmosphere, or Ozone depleting chemicals, or PFAS in the water, microplastics in the ocean. I don't think it's a small list.

CO2 has not ramped up without notice, nor did ozone nor pfas. There are active efforts too handle each of them, and the only reason we aren't cracking down on most of those harder is that the utility they provide to human beings is far greater than the damages right now.

Sure they did. They are now active efforts to handle them, but after they became issues, not while they were ramping up.

Re: Gorilla: Large Language Model connected with massive APIs

#68

Post by fastglass »

ethanbond wrote: post #14

Don't worry folks, the years and years of "of course we'd never connect experimental AI systems to the open internet" assurances were never necessary for safe development and deployment. This reassessment was founded on the last several months that showed us these systems are close to 100% dependable, they never deceive humans, they are entirely aligned with human morals (you know, all those morals we all agree on),…

I like how people talk about this the same way they do gun safety

Re: Gorilla: Large Language Model connected with massive APIs

#69

Post by groby_b »

pixl97 wrote: post #44
bioemerl wrote: post #34

Earlier quoted context omitted.

> What are you basing this on? History. We have had a period of adaptation and experimentation with every other invention in history. This one likely will be no different. > Won't it just be used to augment the amount of harm that humans can do? Sure, but that's everything that makes us more capable. It also makes us more able to defend and create.

>History Right, I learned from Napoleon about the dangers of AI... The large land mammals were doing pretty well too, until a bipedal superintelligence showed up. History is a get thing to learn from and a very poor thing to assume is going to repeat.

If you want to make a claim that we won't even see change coming, maybe the large mammals->humanoids transition isn't the greatest example. That took several hundreds of thousands of years.

They did, to some extent, see it coming (in observing a threat becoming more prevalent and dangerous). But they did not have a way to stop humans. Meanwhile, I'm just staring at the power switch on my computer, wondering how we'd stop AI.

(This is not an argument for unrestrained AI - we should safeguard, absolutely. It is a counterpoint to your "we won't see it coming" argument)

Re: Gorilla: Large Language Model connected with massive APIs

#70

Post by esafak »

robwwilliams wrote: post #55
petters wrote: post #36

Earlier quoted context omitted.

> Maybe this is a good cause to reassess the premise of alignment as a valuable goal? Could you elaborate here? Alignment seems pretty obviously a good thing.

Alignment assumes a well agreed foundational philosophy on what is good, what is fair, what is doable today and tomorrow. Yes, HN contributors might have shared goals for AGI alignment—but we are not the world—-we are a thin slice of one culture.

You can't please everyone, so it is best for good-natured people to get out front. It's the same with any powerful technology.

Are you going to invite religious extremists to the table in the name of fairness?

Post reply on HN